Expedera's packet-based NPU architecture optimizes hardware utilization and reduces DDR memory accesses for AI models, achieving up to 20x throughput improvement and 60% power reduction in smartphone implementations. This technology enables more efficient edge AI deployment, which is crucial for real-time processing in devices that handle streaming content.
Immediate gains in hardware efficiency are critical for moving compute-heavy streaming AI, such as real-time upscaling and video analysis, from the cloud to the device. By solving the memory bottleneck—a primary hurdle for edge devices with battery and thermal limits—Expedera enables sub-millisecond AI responses independent of network stability. This shifts the ecosystem's competitive focus from raw TOPS (tera-operations per second) to sustained performance-per-watt. Watch for smartphone OEMs to prioritize integrated NPU IP that minimizes die size while supporting the 128K token context lengths now standard in mobile-first models like Llama 3.2.
Expedera has built a deployment base of more than 10 million consumer devices using its packet-based NPU IP, with over 200 ExaOps of cumulative compute deployed by customers across multiple leading-edge process nodes. The company's Origin E2 family, which preceded Origin Evolution, was specifically designed for always-sensing camera workloads including 4K video low-light denoising, a use case that directly connects NPU efficiency to streaming-quality video capture on constrained mobile hardware. A leading smartphone manufacturer selected Origin E2 after evaluating every competing NPU IP, achieving a 40x improvement in performance-per-area for 4K video processing tasks.
The Origin Evolution platform extends Expedera's packet architecture into the LLM era with discrete processing blocks for feed-forward, attention, and vector operations that route packets according to network structure and inference phase. For Llama 3.2 1B and Qwen2 1.5B, the company measured more than 75% reduction in external memory moves compared to alternative approaches. This matters for automotive and industrial edge systems where heterogeneous workloads, including perception, driver monitoring, and generative features, must coexist on a single SoC under strict power budgets. Origin Evolution for Edge scales to 32 TFLOPS in a single core and includes a high-speed external memory streaming interface compatible with the latest memory standards.
Expedera's sustained utilization claim of 80% average, compared to what it characterizes as the 20-40% industry norm for tiled AI accelerator engines, positions the company against competitors whose architectures suffer from dark silicon waste. The Origin E2 product page documents that the architecture eliminates memory sharing, security, and area penalty issues faced by lower-performing tiled designs, a claim that resonates with semiconductor buyers evaluating NPU IP for area-constrained smartphone and edge-node designs. The company's co-designed hardware and software stack means models run as trained without retraining or accuracy reduction, reducing integration risk for OEMs.
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source